Data analysis and statistical analysis
The distribution map of common vole samples used in this study (Fig. 1A) was generated using the following R packages: ‘rworldmap’(South, 2011), ‘rworldxtra’(South, 2012), ‘RcolorBrewer’(Neuwirth, 2014), ‘maptools’(Bivand & Lewin-Koh, 2019), and ‘classInt’(Bivand, 2019). Ellipse-like annual relationships between temperature and photoperiod (Fig. 1B, C) were built using ~10-year (between 2000-2019) average monthly ambient temperatures obtained from local weather stations (within 110km of sample location) at http://www.wunderground.com. Photoperiod, based on civil twilight times at dawn and dusk at different locations were retrieved from https://www.timeanddate.com/. Grass growth in spring is used as aproxy for the onset of the favorable reproductive season. Grass growth is initiated at 5-10°C air temperature(Cooper, 1964; Peacock, 1975, 1976). To include all locations in our analysis, a temperature threshold at 6.6°C was used to deduce for further analysis the corresponding predicted critical photoperiod (pCPP) that would initiate optimal timing of reproduction.
The ‘Phyre2’ web portal for protein modeling was used to predict the TSHR protein 3D structure (Fig. 2D)(Kelley, Mezulis, Yates, Wass, & Sternberg, 2015). SNPs were detected by sequence alignments using ‘CLC Sequence Viewer’ (version 8.0) (QIAGEN, Aarhus, Denmark). Chromatograms were checked for sequencing quality and heterozygosity of SNPs in the Mac OS software ‘4-peaks’ (Nucleobytes, Aalsmeer, the Netherlands). Variation in DNA sequences were classified as SNPs if >3 of the specimens contained the mutation. Putative transcription factor bindings sites were predicted using AliBaba2(Grabe, 2002). To statistically test gene-environment associations, we used a population-based approach, in which an environmental variable was modelled as a linear function of population allele frequency(Rellstab, Gugerli, Eckert, Hancock, & Holderegger, 2015). Pearson’s correlation tests were carried out: pairwise distances of allele frequencies were correlated to pairwise geographical distance, pairwise latitudinal difference, pairwise longitudinal difference, pairwise altitudinal difference, and pairwise critical photoperiod difference.P -values were adjusted according to the Benjamini-Hochberg procedure(Benjamini & Hochberg, 1995; Yekutieli & Benjamini, 1999), which is one of the strongly recommended method to use in environmental association analysis(Rellstab et al., 2015). Pairwise linkage disequilibrium heatmaps (Fig. S3) were generated using the R-package ‘LDheatmap’(Shin, Blay, McNeney, & Jinko Graham, 2006). The constructed phylogenetic tree (Fig. S2) from SNP frequency data by using the neighbor-joining method(Saitou & Nei, 1987) was generated using ‘POPTREEW’(Takezaki, Nei, & Tamura, 2014). All other analyses were performed using ‘RStudio’ (version 1.2.1335) and figures were generated using the R-package ‘ggplot2’(Wickham, 2016).